REBOOT(8) REBOOT(8) NAME reboot - bootstrapping procedures SYNOPSIS /etc/reboot [ -s ] [ -n ] [ -a ] [ -d ] /etc/halt [ -n ] DESCRIPTION UNIX is started by placing it in memory at location zero and transferring to zero. Since the system is not reenterable, it is necessary to read it in from disk each time it is to be bootstrapped. Rebooting a running system The preferred way to reboot is to log in on the console, invoke kill 1 to take the system to single user, unmount filesystems with /etc/umount -a and halt and restart the system as described below under Console boots. If access to the console is difficult, /etc/reboot may be used to restart a running system. It normally syncs the disks (see sync(8)), then causes a system boot and an auto- matic disk check. If all this succeeds, the system is brought up multi-user. The options to reboot are: -s Come up single-user after the reboot, without checking the disks. -n Don't sync the disks before performing the reboot. -a Ask on the console for the name of the file to be booted. -d Write a crash dump to the swap area before rebooting. /etc/halt syncs the disks and stops the CPU. If the -n option is enabled, the disk sync is not performed. Power fail and crash recovery The system will normally reboot itself at power-up or after crashes if the auto-restart is enabled on the machine front panel. An automatic consistency check of the file systems is performed. Unless this fails the system will resume REBOOT(8) REBOOT(8) multi-user operations. Console boots: VAX-11/750 Sync the disks if necessary and possible. To recover con- trol of the console (normally it is running as an ordinary Unix terminal), type a control-P. This will halt the CPU and yield a `>>>' prompt from the VAX console subsystem (sic). (Under undocumented conditions, this may fail. If so, hit the white button on the front panel.) The command >>> B will cause a normal automatic reboot. Use >>> B/3 to come up single-user. This also will prompt (with `:') for the name of the file to boot. The format of the file name is disk(m, n)file where disk is `ra' for UDA50 disks with 4KB filesystems, `sa' for UDA50 disks with 1KB filesystems, `up' for UNIBUS SMD disks, `hp' for Massbus SMD disks; m is the drive number (usually 0); n is the file system number on the drive (usu- ally 0); and file is the Unix name of the file on the disk. If the root disk is curdled, the bootstrap program /boot can be read from another disk, say RA drive 1, by >>> B/3 DUA1 To the console subsystem, the RA's are called DU; the Emulex disks are EM. Note that regardless how a system is boot- strapped, it will use drive 0 for the root disk. Console boots: VAX-11/780 To halt a 780, type control-P, then issue the command >>> H to halt the CPU. Rebooting a 780 is like a 750 except for the format of the B command. >>> B REBOOT(8) REBOOT(8) initiates automatic reboot. >>> B ANY comes up single user. Replace `ANY' by `HPS' for hp disk single user reboot, `RPS' for rp reboot, and `UPS' for ra reboot. Generating system core images If the system crashes, it attempts to save a copy of its physical memory to the swap area, to be copied to the file system after reboot (see savecore(8)). To save a core image for a hung system, type on the console (after control-P if necessary): >>> I >>> S C00 This resets the bus and jumps to the start of the dumping code. The system will reboot automatically. Examining system core images Savecore saves in the dump directory two files: the copy of physical memory, called vmcore.n, and a copy of the kernel binary unix.n, where n is a sequence counter. To recover a stack traceback at the time of the crash, type $ adb unix.n vmcore.n $<crash $c Pi(9.1) can examine kernel core dumps. FILES /unix default Unix kernel binary /boot system bootstrap SEE ALSO fsck(8), init(8), rc(8) BUGS Our homemade boot ROM for Emulex disks always boots from drive 0, no matter what you tell it.